Bioenergy is defined as energy obtained from biomass or its metabolic by-products (Calvin et al., 2021). Taking into consideration current global warming, climate change and energy security issues, it has important advantages over conventional fossil fuels (Prasad et al., 2021; Rather et al., 2022). Sweet sorghum [Sorghum bicolor (L.) Moench], as a first-generation energy crop (Yadav et al., 2019), offers a valuable potential source of sustainable bioenergy production (Olivera et al., 2020) due to its high water-use efficiency and resistance to heat and drought conditions (Abebe et al., 2020). Sorghum is also considered a “failsafe” in agroecosystems (Paterson, 2008), due to its resilience and productivity in conditions that are too hot and dry for most other cereals. Sweet sorghum has drawn attention not only for the production of bioenergy from directly fermentable sugars contained in its stalk, but also for its high biomass production capacity (Smith et al., 1987; Murray et al., 2009). Compared to other bioenergy sources such as sugarcane, sweet sorghum has a comparatively low input requirement, high water use efficiency, drought resistance and high yield, and furthermore also produces grain that can be used in food or feed; therefore, it is considered a better alternative for bioethanol production from stem sugars, as both energy and food can be produced simultaneously (Reddy et al., 2005).

Selection for bioenergy producing sorghum cultivars depends on the characterization of biomass-related traits like plant height, sugar content, and fresh biomass. Plant height is an important part of plant architecture (Guden et al., 2020a) that has a significant correlation with biomass production (Calvino and Messing, 2012). Because quantitative trait loci (QTL) for biomass have been identified to be co-localized with height QTLs (Brown et al., 2008), breeders usually target taller genotypes in their sorghum biomass development strategy. The Brix value, which also has a positive correlation with height (Wang et al., 2020a), is an indicator of the total sugar content in stem juice and so represents an important bioenergy trait. Extractable juice content, which can be directly fermented into ethanol, is another trait that contributes greatly to high bioenergy production (Takaki et al., 2015). Considering all these factors, recent breeding studies have focused on crop improvement by identifying and selecting for QTLs that control these traits (Regassa and Wortmann, 2014; Mace et al., 2019). Employing quantitative genetics and molecular-based technologies, several previous mapping studies have been conducted to determine associations between gene/QTL regions and bioenergy-related traits (Kajiya-Kanegae et al., 2020; de Souza et al., 2021). de Souza et al. (2021) used 272 RIL plants from a cross between two sweet sorghum lines (Brandes × Wray) to create a genetic map and mapped 33 QTLs for bioenergy traits, 14 of which were found in all three environments tested. Jin et al. (2021) constructed a genetic map and identified a total of 43 QTLs for biomass traits, several of which were stable across 2 years. However, the high variability of agronomic traits between different locations and years demands further studies, in order to identify stable QTLs that consistently improve traits in multiple environments and genetic backgrounds (Chen et al., 2020).

Single Nucleotide Polymorphism (SNP) markers have higher coverage in plant genomes than other marker types such as RFLP and SSRs (Kumar et al., 2009) and can be identified rapidly using next generation sequencing methods such as whole-genome sequencing, genotyping-by-sequencing (GBS), restriction site-associated DNA sequencing (RAD-seq), double digest restriction site-associated DNA sequencing (ddRAD-Seq), and specific-locus amplified fragment sequencing (SLAF-seq). Among these, ddRAD-seq is a popular and cost-effective method for SNP discovery that includes digesting the DNA with two defined restriction enzymes, pseudo-random sampling, and size-selection of fragments for sequencing (Peterson et al., 2012). It is inexpensive compared to whole-genome sequencing, and as many restriction sites are conserved between different genotypes, no reference genome is required for bioinformatic analysis (Andrews et al., 2016). The application of ddRAD-seq has been reported in many species, such as apple (Malus domestica Borkh.) (Ma et al., 2017), orchid (Orchis spp.) (Roy et al., 2017), onion (Allium cepa L.) (Lee et al., 2018), sesame (Sesamum indicum L.) (Basak et al., 2019) and hazelnut (Corylus avellana L.) (Oztolan-Erol et al., 2021). However, ddRAD-seq has not been applied in sorghum to date.

The objectives of this study were to (i) construct a genetic map for sorghum using ddRAD-seq, (ii) identify QTLs for bioenergy-related traits, and (iii) develop user-friendly PCR-based SNP markers and validate them on diverse genetic sources.

The ddRAD-Seq approach, a flexible diagnostic method for SNP-based genotyping, enabled the construction of a genetic map for an F₂ population created by crossing high-sugar yielding (Erdurmus) and low-sugar yielding (Ogretmenoglu) sorghum genotypes. The obtained SNPs were distributed throughout the sorghum genome indicating the effectiveness of this technique, which made it possible to identify a large number of polymorphisms at a low cost (Peterson et al., 2012). The ddRAD-Seq technique has recently been utilized successfully in a variety of plants, including foxtail millet (Fukunaga et al., 2022), Brassica napus (Chen et al., 2017; Scheben et al., 2020), and sesame (Yol et al., 2021). To our knowledge, this study is the first to demonstrate that ddRAD-Seq is a suitable approach to detect variants at a genome-wide scale in the sorghum genome. We discovered 1438 SNP markers dispersed over 10 LGs ( Figure 1 ), perfectly matching the number of chromosomes in S. bicolor (n = 10), demonstrating the power of the method. Despite the average interval distance being sufficient for precise mapping (Wang et al., 2020b), there were a few> 10 cM gaps in our map, especially on LG7 ( Figure 1 ). Such gaps are occasionally observed in linkage maps in cases such as the mapping population being generated from two parents whose genetic backgrounds are similar in some regions, or due to divergence from the reference genome (Wang et al., 2020b). The total length of our map was 1349.72 cM, while the shortest and longest LGs were 77.46 and 176.36 cM respectively. Our results were compared with those by de Souza et al. (2021), who developed genetic maps for sorghum with GBS, which were similar in length (1368.8 cM) but with greater variance in SNP distribution across the LGs. A comparable GBS method was also used by Somegowda et al. (2022) to construct a genetic map of 1269.9 cM in length, with an average distance of 2.22 cM between adjacent markers. Jin et al. (2021) constructed a high-density genetic map with a length of 1191.7 cM with the single-digest RAD-seq method and the average distance between markers was 1.11 cM. In summary, the linkage map produced here is of very similar length to previously published maps for sweet sorghum, but slightly better in marker distribution and density.

One notable difference from the current work is that all of these previous studies used RILs, whereas the current genetic map was generated from the F₂ lines; different computational approaches were also used in each case. In a comparable study to ours, using ddRAD-seq in a Brassica napus F₂ population, the resulting linkage map (3981.3 cM) was noted to be significantly longer than maps generated using B. napus RILs (Scheben et al., 2020). It was suggested that the heterozygosity of F₂ lines can lead to genotyping errors, inflating the length of LGs; however, in our study the map length was consistent with those generated from RILs, suggesting that any such genotyping errors were successfully filtered out during SNP selection and linkage mapping. We did observe some differences in marker order between the genetic map and the reference genome, especially in the peri-centromeric regions, as well as two putative translocations. While the type of mapping population, number of individuals (Li et al., 2017), choice of parental lines and sequencing methods (Yol et al., 2021) might all cause differences in mapping, the high density and consistency of marker order across the large majority of the ddRAD-seq linkage maps presented here make them useful for localizing QTLs (Scheben et al., 2020). Furthermore, the QTLs identified below were all in regions for which the marker order was closely conserved with the reference genome, indicating that they are likely to be consistent with other sorghum genotypes.

A total of ten QTLs related to bioenergy traits were determined in our population, and PVEs ranged from 8.6% to 35.2% across the environments. Three major QTLs (qPH1.1, qPH7.1 and qPH9.1) associated with PH were found on chromosomes 1, 7, and 9. These QTLs had additive effects that ranged from 13.64 to 29.33 and dominance effects that ranged from -2.23 to 14.36. Previously, four dwarfism genes influencing PH in sorghum, referred to as Dw1, Dw2, Dw3, and Dw4, were described (Quinby and Karper, 1954; Upadhyaya et al., 2013). GA2-oxidase is encoded by the gene Dw1 (Sobic.009G230800; Chr09:57093313–57095643) (Yamaguchi et al., 2016); overexpression of GA2-oxidase lowers active GA levels and shortens plant height (Wang et al., 2020c). Dw2 (Sobic.006G067700; Chr06:42803037–42807134) encodes a protein kinase that is homologous to KIPK, a member of the AGCVIII subgroup protein kinase family which affects endomembrane function and cell division during the formation of sorghum internodes (Hilley et al., 2017; Olivera et al., 2020). The Dw3 (Sobic.007G163800; Chr07:59821905–59829910 (Multani et al., 2003) gene is located on chromosome 7 and encodes an ABCB1 auxin efflux transporter. Other recently reported dwarfism genes include Dw5, found in a mutagenesis library of BTx623 (Chen and Laza, 2019), which affects sorghum plant height similarly to Dw1-Dw4 and decreases plant height by 36-53% (Habyarimana et al., 2020), and similarly Dw7a (Sobic.007G137101) can reduce sorghum height by 40–50% (Hashimoto et al., 2021). However, no protein-coding gene has yet been identified for Dw4 (Hashimoto et al., 2021). In present study, the SNP identified at Chr7_59610235 on QTL qPH7.1 had the largest contribution (34.8%) to PH and is located 220 kb away from the Dw3 a priori candidate gene (Multani et al., 2003). A second putative SNP at Chr7_59803007 was identified 30 kb away from Dw3 ( Table 8 ). Various studies have also presented QTLs controlling plant height in sorghum near this location (Kajiya-Kanegae et al., 2020; Faye et al., 2021) ( Table 8 ). Moreover, we found that the major QTL associated with plant height on chromosome 7, located between 83.204–107.873 cM, overlapped with two smaller QTLs for other bioenergy-related traits, Brix, and fresh biomass weight. This region also contains the Dw3 gene, and while it is possible that this gene also regulates biomass through limitation of plant height, the presence of the additional QTL for Brix suggests that this region may be a hot spot for genetic control of bioenergy production in sorghum. We also mapped qPH9.1 (Chr9_3610075) on chromosome 9 across both environments, but this QTL is distinct from the previously reported Dw1 gene (Sobic.009G230800, Chr09:57093313–57095643) (Hilley et al., 2016; Yamaguchi et al., 2016). It may demonstrate the presence of another candidate PH gene because QTL overlapping with qPH9.1 were also mapped to this position in a study by Wang et al. (2014). When we evaluated the environments individually, qPHA4.1 was colocalized with previous QTLs (Phuong et al., 2013; Zhang et al., 2015) on chromosome 4, qPHA9.2 was determined to be 33 kb away from Dw1 on chromosome 9 (Hilley et al., 2016).

The stalk sugar yield of sweet sorghum, which is influenced by sugar content and stalk juice, determines its potential for ethanol production (Guan et al., 2011). Although stem juice yield is a key factor in this regard, limited studies have been conducted to identify related genes/QTLs (Murray et al., 2008a; Murray et al., 2008b; Guan et al., 2011). In this study, one major QTL (qPJ6.1) for PJ was determined on chromosome 6, which explained 35.2% of the phenotypic variance. In recent years, a few studies have reported QTLs associated with PJ in sorghum that overlap with this position (de Souza et al., 2021; Ji et al., 2022). Mocoeur et al. (2015) identified a QTL with a high PVE (37.5%) located on chromosome 6 (48787617-50907017) associated with juice yield. Moreover, Burks et al. (2015) also determined a QTL associated with juice yield, and this QTL had a significant association with midrib color, sugar yield, sugar yield by weight, moisture, juice volume, and juiciness. These two QTLs also were found to colocalize with our QTL, qPJ6.1, on chromosome 6. Moreover, the qPJ6.1 QTL for PJ was located at 267 kb from the D-gene (Sobic.006G147400, which encodes a plant-specific NAC transcription factor that controls the juiciness of the stem (Zhang et al., 2018). Surprisingly, the positive effect on juice content at qPJ6.1 was associated with the allele from the ‘low-juice’ parent (Ogretmenoglu). This reflects the fact that juice yield is a complicated quantitative trait linked to plant height, biological yield, and sugar content (Burks et al., 2015; Han et al., 2015); it can be assumed that one of these other traits masks the effect of the beneficial qPJ6.1 allele in Ogretmenoglu. We also mapped four major QTLs associated with FBW; one of them, qFBW6.1 (Chr6_39024006), had a PVE of 14.5% and can be considered a novel QTL, because it did not overlap with any previously determined QTLs on chromosome 6.

Common QTLs determined across more than one environment or population are defined as stable QTLs, which may be more useful in breeding (Qu et al., 2021). Using variance analysis, the G × E interaction is determined as a QTL-by-environment (Q × E) interaction in QTL mapping studies (Chen et al., 2010). In the present study, variance analysis revealed that environmental effects play an important role in the development of bioenergy-related traits, especially for PH. Ten QTLs in total and three QTLs for PH were identified from combined data from two environments, including both lowland and highland sites. During the vegetative period, the lowland site had a much higher maximum temperature (39.7-40.9°C) than highland (33.0-34.9°C), and very little rain ( Table 1 ). In this period, the optimal temperature range for sorghum is 26-34°C (Hammer et al., 1993). This situation may have caused variation in bioenergy-related traits; therefore, the QTLs detected across both environments are more likely to be stable under different conditions, although this should be confirmed with additional phenotyping data from different years. This was supported by the finding of two plant height major QTLs (qPHA4.1 and qPHA9.2) that were only evident when data from the two environments were analyzed individually, and do not co-localize with those determined using combined data ( Figure 2 ).

Phenotypic traits are highly influenced by both environmental changes and growth stage. Therefore, molecular markers should be integrated into breeding programs to allow desired traits to be selected rapidly and independently of environment effects. In the present study, CAPS markers closely linked with several QTLs for pH, BX, PJ, and FBW traits were developed. Two markers, PH1 and BX3, were positioned in the same region of chromosome 7 because of their overlapping QTLs. PH1, PH7, BX3, BX6, PJ13, PJ15, and FBW5 markers were each strongly linked to their respective QTL regions. Moreover, the PH1 marker was approximately 220 kb from the Dw3 gene, while FBW5 was derived from the novel QTL (qFBW6.1) for FBW identified on chromosome 6. All markers, which were validated on eight lines and three cultivars, showed perfect matches with the detected SNPs. Although CAPS markers have been widely used in genetic analysis, there is only one study in which genome-specific CAPS markers have been developed in sorghum. Burow et al. (2019) developed CAPS markers to identify the genotype of bmr6-12 lines and determined that the GA 340 cultivar was classified as bmr6 while BTx623-bmr and B03288bmr were classified as bmr12. Several SNP markers closely linked with the eight major and two minor QTLs reported here can supply beneficial information to facilitate fine mapping of bioenergy-related traits. CAPS markers converted from these SNPs could also be readily used for pyramiding genes for bioenergy-related traits in breeding programs in the near future.

We used the ddRAD-seq approach to construct a high-density linkage map developed from an F₂ population in sweet sorghum, and to screen the linkage map for bioenergy-related traits. Ten QTLs associated with plant height, Brix, fresh biomass weight, and plant juice were identified from combined data from two contrasting environments, including lowland and highland sites. In addition to the identified major QTLs, a novel QTL for effective fresh biomass weight and newly improved CAPS markers, one of which was from this region, add to the current knowledge about these traits and provide additional capacity for the improvement of bioenergy production in sorghum.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi.nlm.nih.gov/, PRJNA894980.

BU designed and supervised the experiments. EY, BU and SL contributed final editing of manuscript. BU analyzed the data of map and QTL. BG, EY, and BU conducted the experiments and analyzed the data. BG, EY, CE and BU conducted the field trials. All authors contributed to the article and approved the submitted version.